MEBI 591B Public Health Informatics Colloquium © Phil Hurvitz, 2006 Measuring Physical Activity and Location in Real Time Phil Hurvitz University of Washington College of Architecture and Urban Planning Urban Form Lab gis.washington.edu/phurvitz MEBI 591B Public Health Informatics Seminar

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 2 (of 45) Confidentiality Unpublished data Please do not distribute

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 3 (of 45) Overview Introduction/Background/Relevance What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 4 (of 45) Overview Introduction/Background What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 5 (of 45) Introduction/Background: Obesity Obesity threatens personal health and may bankrupt the US health care system Obesity incidence has increased dramatically over the last 20 years Source: CDC BRFSS (

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 6 (of 45) Introduction/Background: Obesity Trends

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 7 (of 45) Introduction/Background: Diet & Exercise Nutrition guidelines: “Eat more grains, fruits, vegetables…” Health care system says, “Eat less, exercise more.” Technology and food provides choices that are not conducive to healthy lifestyles

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 8 (of 45) Introduction/Background: Physical Activity Increasing physical activity is important for maintaining or decreasing weight, and for general health The built environment can either promote or hinder physical activity, e.g., Presence/absence of sidewalks Presence/absence of utilitarian destinations (e.g., restaurants, retail stores, restaurants, banks) Research Question: How does physical activity vary with different compositions and configurations of environment?

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 9 (of 45) Overview Introduction/Background What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 10 (of 45) Introduction to GIS: What is GIS? A computer-based method for Capture, Storage, Manipulation, Analysis, and Display of spatially referenced data

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 11 (of 45) Introduction to GIS: What is GIS? Any object or phenomenon that is or can be placed on a map can be stored, managed, and analyzed in a GIS. Built environment features (streets, buildings, bus routes, restaurants, schools) Households (address points, tax-lot polygons) Individuals (points or travel lines/polygons) Ground surface elevation or slope Movement of objects through time and/or space Demographics, socioeconomics Patient residence, work, and school locations Exposure or risk estimation Disease occurrence

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 12 (of 45) Introduction to GIS: Data Framework GIS combines coordinate (map) and attribute (tabular/statistical) data

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 13 (of 45) Introduction to GIS: Coordinate Framework

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 14 (of 45) Introduction to GIS: Address Location GIS can match address records to spatial location

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 15 (of 45) Introduction to GIS: Analysis Analytical techniques (a very simple list) Spatial aggregation Disease rates per census or zip code area Buffering How many pedestrian-auto collisions within 1 mile of schools? Overlay/Proximity analysis How much of each census block group is affected by a toxic aerosol plume? How many parcels of each type of land use are within ½ mile of all walking locations visited within a day? Surface generation, interpolation Trend or density surfaces Kriging

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 16 (of 45) Introduction to GIS: Risk or Exposure Estimation Miranda, M. L. and D. C. Dolinoy Neurotoxicology. 26(2)

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 17 (of 45) Introduction to GIS: Risk Surface Estimation Kernel density estimator (KDE) creates a Gaussian surface for each individual point location and sums each individual surface across XY space

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 18 (of 45) Introduction to GIS: Risk Surface Estimation Fast food restaurant KDE

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 19 (of 45) Introduction to GIS: Risk Estimation Is there a relationship between fast food density and obesity? p-value = 0.155

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 20 (of 45) Introduction to GIS: Risk Surface Estimation Kriging (geostatistical analysis) Hellstrom, L., L. Jarup, B. Persson and O. Axelson J Expo Anal Environ Epidemiol. 14(5) sig. relationship between Pb in soil and blood ♀ eating homegrown vegetables

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 21 (of 45) Overview Introduction/Background What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 22 (of 45) Measuring Physical Activity: How? Subjective Observation Self-Report Stanford 7-Day Activity Survey International Physical Activity Questionnaire (IPAQ) Travel Diaries Objective Pedometers Accelerometers New Generation Devices

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 23 (of 45) Measuring Physical Activity: Benefits & Drawbacks TypeBenefitsDrawbacks Subjective Observationdoes not require effort on part of subject accuracy varies by observer & instance high cost Self-Reportdoes not require observer low cost over-reporting common recall bias Objective Pedometerlow cost easy to use acceptable for free-living subjects not suitable for all populations no activity discrimination no location no temporal resolution Accelerometerno activity discrimination no location New Generation Devices varies

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 24 (of 45) Measuring Physical Activity: New Generation Devices Intelligent Device for Energy Expenditure and Activity (IDEAA) sensors attached to skin (cumbersome) relative accelerometry of different body parts no locational capability no external environmental cues $4,000 per unit

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 25 (of 45) Measuring Physical Activity: New Generation Devices IDEAA: recognizable activities

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 26 (of 45) Measuring Physical Activity: New Generation Devices IDEAA: categorized activities by time

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 27 (of 45) Measuring Physical Activity: New Generation Devices Multi-Sensor Board UW/Intel invention, recent development single sensing unit with data logger (smart phone) easily worn measures multiple environmental data streams obtains XY locational data estimated $100 per unit cost in large manufacturing run

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 28 (of 45) Measuring Physical Activity: New Generation Devices Multi-Sensor Board On-board sensors: accelerometry audio IR / visible light high-frequency light barometric pressure humidity, temperature geophysical location (from GPS) Multivariate data stream can be interpreted as a number of common activities using Hidden Markov Model with Decision Stumps classifiers Used in ECOR Pilot & Feasibility Study

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 29 (of 45) Measuring Physical Activity: New Generation Devices Multi-Sensor Board Activity Classifier (overall accuracy > 95%) Validated against videography SittingStandingWalkingJogging Walking up stairs Walking down stairs Riding a bicycle Driving car Riding elevator down Riding elevator up Sitting 89.8%38.5%0.5%0.4%33.4% Standing 10.1%50.8%1.4% Walking 0.1%7.4%97.7%5.2%2.5% Jogging 100.0% Walking up stairs 94.8% Walking down stairs 0.5%97.5% Riding a bicycle 3.3%99.6% Driving car 66.6% Riding elevator down 100.0% Riding elevator up 100.0% Classified Activity (by HMM) Precision Labeled Activities

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 30 (of 45) Measuring Physical Activity: New Generation Devices Multi-Sensor Board Classification of Activity 90-minute interval

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 31 (of 45) Overview Introduction/Background What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 32 (of 45) Measuring the Built Environment: What and Where? What to Measure? Based on Research Question(s) GIS Data Sources Point Locations Buffer Measures Proximity Measures Where to Measure? Home-centered Frank et al Moudon et al Where does activity take place in real time?

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 33 (of 45) Measuring the Built Environment: A GIS Based Approach Point-centered Analysis of Location Any number of different data sets can be quantified Enumeration & relative proportion of different land uses Parcel density Street-block size Total length of sidewalk Number of intersections, lighted crosswalks Area and count of parks Distance to different built environment features We should quantify & analyze all locations that are experienced during the day, not only the home location Work & school environments may be key determinants of physical activity

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 34 (of 45) Measuring the Built Environment: A GIS Based Approach

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 35 (of 45) Measuring the Built Environment: A GIS Based Approach GIS analysis results for each location buffer (count) measures proximity measures

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 36 (of 45) Overview Introduction/Background What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 37 (of 45) RRF Funded Research Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space MSB to capture Activity type Location Walkable-Bikeable Communities GIS Software Quantifying & analyzing the Built Environment

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 38 (of 45) RRF Funded Research: Analysis Plan MSB activity & location Validity tests against diary (real-time location & activity), IPAQ (self-reported physical activity summary) WBC location analysis of Built Environment Data overload? 15 h * 60 min/h * 60 s/min * 7 d * 40 subjects = 15,120,000 data points

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 39 (of 45) RRF Funded Research: Analysis Plan Sampling strategy for data reduction without loss of variability 10% sample → 1.5 million data points (time or distance?)

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 40 (of 45) RRF Funded Research: Analysis Plan This will be the first study to measure objectively both physical activity types and Built Environment in a real-time, real-world setting with free-roaming individuals Statistical associations? Activity types/intensities & Built Environment types? What do we gain if a pattern is discovered? Policy recommendations Quantitative urban design guidelines A new “gold standard” for measurement of physical activity in real-time

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 41 (of 45) RRF Funded Research: Results from Pilot & Feasibility Study Sample demographics

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 42 (of 45) RRF Funded Research: Results from Pilot & Feasibility Study MSB activity & location

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 43 (of 45) RRF Funded Research: Results from Pilot & Feasibility Study Automatic classification vs. self-report (42 diary entries) p=0.05, Fisher’s exact test * “None” indicates the classifier was not able to classify a given activity † “Shopping” was a user-added activity type that had no match in the automatic classification scheme

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 44 (of 45) Overview Introduction/Background What is GIS, and what is its role in Public Health? Measuring Physical Activity Measuring the Built Environment UW-RRF Funded Research: Validation of New Technologies and Methodologies for Measuring Physical Activity and Location in Real Time-Space Analysis Plan Suggestions/Questions

MEBI 591B Public Health Informatics Seminar © Phil Hurvitz, 2007Slide 45 (of 45) Suggestions/Questions Phil Hurvitz gis.washington.edu/phurvitz gis.washington.edu/phurvitz/msb